Swarm Herding: A Leader-Follower Framework For Multi-Robot Navigation
This addresses efficient navigation and coverage for multi-robot systems, but it appears incremental as it builds on existing leader-follower and deformable object models.
The paper tackles multi-robot navigation for large teams by proposing a leader-follower framework where leaders corral followers, resulting in a decentralized control strategy that enables coverage over a time-varying domain defined by leader positions.
A leader-follower framework is proposed for multi-robot navigation of large scale teams where the leader agents corral the follower agents. A group of leaders is modeled as a 2D deformable object where discrete masses (i.e., leader robots) are interconnected by springs and dampers. A time-varying domain is defined by the positions of leaders while the external forces induce deformations of the domain from its nominal configuration. The team of followers is performing coverage over the time-varying domain by employing a perspective transformation that maps between the nominal and deformed configurations. A decentralized control strategy is proposed where a leader only takes local sensing information and information about its neighbors (connected by virtual springs and dampers), and a follower only needs partial information about leaders and information about its Delaunay neighbors.